Abstract
ABSTRACT. Autologous arteriovenous fistulas (AVF) have the best 5-yr patency and the lowest complication rate among hemodialysis vascular accesses. However, maturation requirements to optimize survival are unknown. A longitudinal cohort study was conducted to ascertain risk factors for failure, maturation time, and survival of the first AVF. All patients who initiated hemodialysis between January 1, 1997, and December 31, 2002, in three centers were included in this study. Analysis was restricted to patients who received an AVF. Cox regression was used to estimate the association between predictors of interest and primary and secondary AVF survival. Of the 535 patients enrolled (mean age, 66.5 yr; 57.8% male; 26.7% diabetic), 513 (96%) received an AVF. Patients who initiated with catheters (47%) cannulated their AVF earlier (median maturation period, 0.78 versus 1.80 mo; P < 0.001). Median primary and secondary survivals were longer than 50 and 72 mo, respectively. After adjustment for confounding factors, cardiovascular disease (hazard ratio [HR], 1.84; 95% confidence interval [CI], 1.26 to 2.67), utilization earlier than 1 mo after placement (HR, 1.94; 95% CI, 1.34 to 2.82), and referral within 3 mo of dialysis start (HR, 1.55; 95% CI, 1.04 to 2.32) were associated with a reduction in primary AVF survival. Presence of cardiovascular disease (HR, 2.21; 95% CI, 1.38 to 3.55), maturation time <15 d (HR, 2.12; 95% CI, 1.20 to 3.73), and presence of catheters at hemodialysis initiation (HR, 1.79; 95% CI, 1.13 to 2.84) were associated with lower secondary AVF survival. It is concluded that cardiovascular disease, late referral, temporary catheters, and early cannulation are associated with impaired AVF survival. It is recommended that AVF be allowed to mature at least 1 mo before cannulation. E-mail: p.ravani@libero.it
Creation and maintenance of a well-functioning vascular access (VA) remains one of the most challenging problems in delivering adequate hemodialysis (HD) therapy. As compared with arteriovenous grafts and permanent cuffed central venous catheters, native arteriovenous fistulas (AVF) are viewed as being superior, because of the much smaller number of procedures associated with use and longer overall survival (1). However, AVF are not readily utilizable after placement, and early cannulation has been shown to be associated with shorter survival (2). Indeed, the optimal maturation period to make cannulation easier and improve AVF survival is not established, although it is common practice to delay utilization after insertion to allow enlargement and “arterialization” of the vein. This maturation process is expected to result in adequate blood flow through the fistula and in lower risk of local hemorrhage at the cannulation site, wall damage, fibrosis, and final occlusion. The current National Kidney Foundation’s Dialysis Outcomes Quality Initiative guidelines endorse this practice and recommend that initial use be delayed for at least 4 wk after surgery, although this guidance is based on opinion (3), and a shorter maturation time has been suggested to be sufficient by a recent large observational study (4).
In a previous study by our group, we reported that the presence of temporary catheters at dialysis initiation was associated with earlier cannulation and shorter survival of the first AVF created for incident patients, suggesting that the insufficient maturation period may be a causal link between catheter utilization and failure (5). In the present study, we sought to confirm and expand on those findings by including three large dialysis centers with a similar approach to VA surgery and further explore the interrelationship among timing of AVF creation, maturation before utilization, and survival.
Materials and Methods
Subjects
Data for this longitudinal cohort study were collected by means of a computerized database containing demographic and clinical information on all consecutive ESRD patients who were older than 18 yr, receiving a new AVF, and entering maintenance HD treatment programs at three dialysis units in northern Italy (Cremona, Brescia, and Lodi) from January 1, 1997, to December 31, 2002. These files were continuously updated and included date of AVF surgery referral and creation, type of access inserted, placement location, first use and failure, and all access-related complications and procedures. Data for these analyses were restricted to patients who received the VA placement for the first time and by one of the local renal physicians in charge of the VA-related procedures.
Vessel Choice, Fistula Cannulation, and Catheters
All patients in this cohort underwent a preoperative clinical examination of arm arteries and veins using out-flow occlusion by means of a tourniquet. Vascular mapping of the arm was performed regularly before surgery when the patient had previously undergone superior cava vein catheterization, the patient had a previous AVF dysfunction (in the case of revisions), or no vein was apparent. An echo-power Doppler was supported by a venogram, if indicated. Site and type of VA were chosen mainly on the basis of clinical examination, with no minimum criteria for artery and vein diameter, and all interventions were performed under local anesthesia. To increase the use of native AVF, we prepared a standard autogenous AVF at the distal lateral wrist (radiocephalic anastomosis) and the antecubital fossa of the elbow (radiocephalic and brachiocephalic or, less frequently, brachiobasilic anastomoses), starting distally with the nondominant arm to preserve proximal sites. The prescribed interval time before cannulation was 2 to 4 wk. The first cannulation was performed by a nephrologist only when it was considered necessary by the nurses in charge. Patients without a functioning AVF started dialysis by temporary catheters inserted in both the inferior (femoral vein) and the superior cava vein (internal jugular vein) systems on both sides, avoiding subclavian veins.
Study Variables and Outcome Definitions
Covariates identified a priori as possible risk factors for AVF failure included age at the beginning of HD, gender, and race and all comorbid conditions abstracted from the hospital electronic databases of admissions and discharges. The last were defined on the basis of diagnosis-related group classification and International Classification of Diseases, Ninth Revision. Patients were considered to have cardiovascular disease when they had a documented diagnosis of congestive heart failure (NYHA class II or greater) or ischemic heart disease caused by coronary artery disease with or without myocardial infarction or clinical manifestations of peripheral vasculopathy or cerebrovascular disease (stroke, transischemic attack). Predialysis care was evaluated considering both the referral pattern and the frequency of visits. Use of temporary catheters at HD initiation and their side of placement (either the same or opposite to that of the AVF) were considered independent variables, as well as the interval between creation and use of the first AVF. This maturation period was collapsed into the categories of <15 d (reference), 15 to 30 d, 30 to 60 d, and longer. These cut-off points were used on the basis of recent literature reports (4) and analysis of the actual data.
A VA was considered patent when it worked well after creation or any further intervention and was capable of providing blood flow sufficient to obtain an adequate dialysis dose within a maximum of 5 h/session (Kt/V ≥1.2 according to the Daugirdas second-generation formula). For reaching this goal, the average dialysis blood flow rate provided by a patent AVF was 320 ml/min (280 to 350). VA failure was defined as failure to mature, definitive clotting, or malfunction caused by stenosis or partial thrombosis, and diagnosed as described previously (5). Revisions (restoring patency of the same VA—same artery and vein and same location—by surgery, pharmacomechanical intervention, or angioplasty) were distinguished from new creations, the former without and the latter with a change in VA conduit, e.g., insertion of a new VA in the upper arm on the same side (when the first VA was distal) or in the other arm (6). Accordingly, primary survival (the major outcome measure of the study) was defined as the intervention-free period to first failure, and cumulative (unassisted) primary patency was defined as the relative frequency of all VA at risk at any one time that were still functioning without further interventions. Secondary survival was defined as the time to final failure, regardless of the number of revisions required to salvage, and cumulative assisted secondary patency was defined as the proportion of all VA at risk at any one time that were still functioning, including all revisions. For testing the predictive role of time to cannulation, survival analyses were conducted using the date of first venipuncture as time 0, excluding those AVF that were not used at all. This strategy was adopted for the first time by Rayner et al. (4,7⇓) and avoids superimposing the potential predictor over the outcome measure and violation of the Cox’s model assumptions.
Statistical Analyses
Logistic regression was used to investigate the relationship between patient characteristics and whether AVF were first cannulated in a short time interval (<30 d, early utilization) versus a longer time interval (≥30 d). Analyses were repeated using also 15, 60, and 90 d as early cannulation definitions. Logistic regression was also used to categorize and establish cut-off levels of time to cannulation in relation to failure.
Survival functions were described using the Kaplan-Meier technique. The log-rank test was used for univariable comparisons. Patients were censored when they were changed to PD therapy, were transferred to another dialysis unit, received a kidney graft, died, or had a functioning AVF on the final observation date (December 31, 2002). Cox’s proportional hazards regression was used to model time to event as a function of cannulation time and use of temporary catheters at dialysis start. The potential effect of AVF location (distal versus proximal) was evaluated both testing the effect of this covariate and stratifying the models. Risk factors related to patient demographics and comorbid conditions, the side of the catheter, and departmental organizational issues (e.g., surgeon, referral timing and predialysis care, previous PD therapy) were also considered. All variables used in the equations were chosen a priori and retained in the models when there was biologic plausibility or when univariate analyses suggested that they may be associated with the event or may confound the relationship between the covariate of interest and the event. The proportional hazards assumption was checked for each model by inspection of the complementary log minus log plots.
For both logistic and survival regression analyses, stepwise method was used to obtain the best multivariate model. The −2 Log likelihood ratio (−2 Log L) statistics was used for goodness-of-fit comparisons (8,9⇓). Estimated relative risks (odds ratios for logistic regression and hazard ratios for time to event analyses) and their 95% confidence intervals were calculated with the use of the estimated regression coefficients and their standard error. The contribution of covariates to explain the dependent variable was assessed by means of a two-tailed Wald test, with P < 0.05 considered significant. The P value for variable removal within the multivariate analyses was set to 0.10. All statistical analyses were performed using SPSS software, version 11 (SPSS Inc., Chicago, IL).
Results
Patient Characteristics, Referral Patterns, and Follow-up on Dialysis
A total of 535 consecutive patients (198 from Brescia, 175 from Cremona, and 162 from Lodi) were enrolled in the study and observed over an average follow-up of 42 mo (1872 patient-years). At the time of study enrollment, the mean (SD) patient age was 66.5 (14.2) yr, 98% of patients were white, and 58% were male. The follow-up of the cohort was complete, no patient was referred to another center for VA surgery, and all VA were placed within 6 d of surgical referral. Table 1 presents underlying renal diseases, cardiovascular risk factors, and comorbid conditions. Most patients were hypertensive and had comorbid conditions. Twenty-eight percent of the patients started dialysis within 3 mo of seeing a renal specialist, whereas 24.3% did not receive any predialysis care.
Table 1. Description of renal disease, risk factors, comorbidities, and predialysis carea
VA Surgery, Temporary Catheters, and Time to Cannulation
A total of 513 (96%) patients received an AVF (420 distal and 93 proximal fistulae). A total of 446 patients used their first AVF (32 after a salvage procedure) and represented the main focus of the study. Of the remaining, six died before AVF use and 61 received a new VA as a result of AVF failure before utilization. At initiation of HD therapy, nearly one half (47.2%) of patients used a temporary dialysis catheter and the remainder used an AVF. Patients with catheters at HD initiation cannulated their AVF earlier than those who started using their AVF (median [interquartile range (IQR)] maturation time, 0.78 mo [0.4] versus 1.8 mo [2.3]; P < 0.001). No difference was found in the distribution of renal diseases, cardiovascular risk factors and disease, and comorbid conditions by maturation period shorter and longer than 1 mo (both including and excluding AVF used after salvage). Conversely, the proportion of patients with early cannulation was significantly (P < 0.001) higher among those who were referred within 3 mo of seeing a nephrologist (58.5 versus 39.6%) and those with follow-up absent (61.3 versus 39.6%) and among patients with catheters at HD start (69.1 versus 25.3%). Adjusting for confounding factors, the intensity of predialysis follow-up yielded the best multivariate model possible in the absence of a temporary catheter at HD initiation (Table 2, model A). The latter proved to be the most powerful independent predictor of earlier cannulation in the final model (Table 2, model B). Similar results were obtained considering AVF used after revision and also using 15 and 60 d as the cut-off point for binary outcome definition (data not shown).
Table 2. Risk factors associated with early cannulation (defined as <30 versus >30 d), excluding AVF cannulated after revisiona
Primary Survival
Median primary survival from cannulation of the first AVF was longer than 50 mo, with a mean follow-up of 44.9 mo, 1549 AVF-years at risk, and 7.6% annual failure rate. Primary survival of the first distal and proximal AVF was not significantly different. Factors that predicted worse survival were referral within 3 mo of dialysis initiation (P = 0.004), dialysis start with a catheter (P < 0.0001), age older than 65 yr (P = 0.002), presence of vascular disease (P = 0.003), presence of cardiovascular disease (P = 0.001), and, at a borderline level, presence of heart failure (P = 0.07). No difference was found by gender and other considered prognostic factors, including the side of the catheter. AVF within longer maturation time categories had longer survival both including and excluding early failures (P = 0.0007). Figure 1 depicts survival plots by use of catheters and time to cannulation category. Multivariable analysis showed that the presence of cardiovascular disease, late referral, and a maturation time of <30 d were independent predictors of lower survival probability (Figure 2).
Figure 1. Kaplan-Meier curves of time to failure (primary patency from first cannulation) by use of catheters (CVC) at dialysis start (A) and by maturation category in days (B). Crosses mark point of censor.
Figure 2. Risk factors associated with primary and secondary failure. Hazard ratios plotted using a logarithmic scale. Both Cox models are adjusted for baseline characteristics, center effect, arteriovenous fistula locations, and considered comorbid conditions (except cardiovascular disease [CVD]).
Secondary Survival
Life expectancy of the first AVF after cannulation was longer than 72 mo. The mean follow-up was 54 mo, with 2007 AVF-years at risk and a 4% failure rate. The same factors that predicted worse primary survival were also associated with greater risk of final failure. After adjusting for confounders, the presence of cardiovascular disease, utilization of catheters at HD initiation, and early cannulation were independent predictors also of final failure (Figure 2). Of note, a lower cut-off level of time to cannulation was associated with the greatest risk of final failure. This finding is more fully discussed next.
Discussion
The present analysis in a large cohort of patients who initiated HD shows that late referral and use of temporary catheters at HD initiation predicted earlier utilization of the AVF, which in turn proved to be a strong independent predictor of shorter primary and secondary survival. The analysis shows also that the presence of cardiovascular disease (e.g., history of heart failure, coronary artery disease, peripheral vasculopathy) independently predicted greater risk for AVF failure. The implications of these findings are twofold. First, the goal of planning the creation of an AVF implies not only that this permanent access be present at the patient’s initial HD treatment but also that it be left to mature for more than is commonly done in clinical practice. Second, strict control of cardiovascular risk factors and appropriate treatment of cardiovascular disease during the course of progressive chronic kidney diseases may have beneficial consequences also on AVF maturation and survival, further contributing to the expected benefits in terms of morbidity and mortality of the ESRD population.
The first important finding of our study is the association between late predialysis care and earlier AVF cannulation. The use of temporary catheters proved to be the main link between the two, being associated with more than a sixfold increased odds of earlier AVF venipuncture. Use of femoral or jugular temporary catheters at dialysis initiation was also a strong univariate predictor of shorter AVF survival, as described previously in larger studies (4,7⇓). It is interesting that it was independent of age, gender, diabetes, side of insertion, and other comorbidities but not of time to cannulation. These findings are consistent with the hypothesis that catheter complications anticipate the use of AVF, which in turn increases the risk of failure as a result of insufficient maturation (7). Under this hypothesis, one would expect the impact of the catheter to be confounded with that of its proximate consequence (earlier AVF cannulation), limiting the power of the analysis to detect an independent effect, as proved to be the case.
Considering time to cannulation, our findings are consistent with those recently reported by the Dialysis Outcomes and Practice Patterns Study (4,7⇓). Rayner et al. (4) described a higher risk of AVF primary failure for incident patients who had a previous temporary access and in cases of cannulation within 14 d after creation, with no significant survival advantage associated with longer maturation time. Our results partly confirmed this conclusion, in that AVF cannulated <15 d after creation showed the lowest primary and secondary patencies, and maturation period was retained in the Cox’s model of secondary survival when dichotomized in <15 d versus >15 d. This suggests that no salvage intervention, when feasible, could remedy the detrimental consequences of such an early cannulation. However, the risk of primary failure was 50% less in AVF that were left to mature for 1 to 2 mo and was even lower in cases of a longer maturation period, as compared with cannulation earlier than 15 d. This is the first large study to support the Dialysis Outcomes Quality Initiative guidelines recommendation to allow AVF to mature for at least 1 mo (and ideally 3 to 4 mo) before cannulation (3). It is interesting that in the final primary and secondary survival models, late referral and presence of temporary catheters at dialysis start also proved to be explanatory variables. This may suggest that factors other than shorter maturation time can affect the maturation process of the AVF and its survival. It was reported recently that intimal hyperplasia of the radial artery before AVF creation was associated with a higher risk of early failure of the first AVF in 59 patients who started HD (11). It could be hypothesized that long-lasting anemia and arterial hypertension, as well as additional cardiovascular risk factors, may be associated with peripheral vessels’ abnormalities, potentially affecting AVF survival. Therefore, even in the absence of overt hematomas, early cannulation may not only increase the risk of AVF failure through subclinical microhemorrhages, fibrosis, and vessel wall damage but also interfere with the access maturation process itself. In addition, the presence of a temporary catheter may contribute to a less favorable fistula maturation through changes or losses in biologic factors and greater risk of infection, thrombosis, and venous stenosis. Although tunneled cuffed catheters were not placed in the present cohort at the beginning of dialysis therapy, we cannot exclude that similar complications would be associated with their use. However, permanent silicone catheters present the theoretical advantage of higher patency rates and better biocompatibility and, therefore, may show different effects on time to cannulation and AVF survival.
The final important finding of our study is the negative effect of cardiovascular disease on AVF survival. In the present series, presence of cardiovascular disease conferred an 83% and twofold increase in probability of primary and final AVF failure, respectively, similar to what was recently reported (7). The explanation for this may be complex and multifactorial. Whether the presence of cardiovascular disease has a direct effect on AVF maturation and survival or is simply a marker of severe comorbidity, poor predialysis care, or uncontrolled uremia cannot be ruled out from our database. Worsening cardiac function may result in reduced blood flow to the AVF; in addition, known or unknown uremic factors may exert their influence through affecting cardiac and endothelial function, as well as the arteriosclerotic process itself (12,13⇓).
The major strengths of the present work are its relatively large size, its prospective design and follow-up, and its comprehensive assessment of outcomes, although the role of specific markers of uremic complications (e.g., divalent ion metabolism control, hemoglobin and serum albumin levels) have not been evaluated, and some variables such as smoking habit and hypertension have not been studied in detail. Several limitations deserve mention. Maturation was not defined by diagnostic objective tools (14), and the role of potential risk factors such as the skill of the dialysis staff in cannulation (15), occurrence of local hematomas at the cannulation site, and measures of dialysis adequacy as well as dialysis-related hypotensive episodes were not tested. In addition, the potential association between poor maturation and early cannulation cannot be ruled out from our database, although it would be unlikely to occur in clinical practice. None of these limitations, however, invalidate our findings that shorter maturation time and the presence of cardiovascular disease, when present as defined in the study, were associated with a significant reduction in AVF survival. Finally, as is true of any observational design, the causal inferences drawn from the data should be considered hypotheses and ideally should be proved in experimental studies. Therefore, we acknowledge that all risk factors identified in the present study may just be proxies for as-yet-unidentified factors that more directly account for AVF failure. However, whether a cause or a marker for a cause, a modifier or even a confounder, a risk factor remains a predictor of event occurrence (16).
In conclusion, a maturation period shorter than 1 mo and the presence of cardiovascular disease seem to be major predictors of AVF failure. Therefore, further efforts should be made to obtain earlier patient referral to a nephrologist, target and control cardiovascular risk factors, and place AVF timely, avoiding the use of catheters at HD initiation. Although AVF may potentially be inserted in >90% of incident HD patients, a longer maturation time is likely to be necessary to reduce the VA-related complications and procedures.
- © 2004 American Society of Nephrology
Jump to section
More in this TOC Section
Cited By...
- Definitions and End Points for Interventional Studies for Arteriovenous Dialysis Access
- Outpatient Nephrology Referral Rates after Acute Kidney Injury
- Outcomes of Arteriovenous Fistula Creation after the Fistula First Initiative
- Modeling Survival of Arteriovenous Accesses for Hemodialysis: Semiparametric versus Parametric Methods
- Hemodialysis Arteriovenous Fistula Patency Revisited: Results of a Prospective, Multicenter Initiative
- Comorbidities Do not Influence Primary Fistula Success in Incident Hemodialysis Patients: A Prospective Study
- Fistula First Initiative: Advantages and Pitfalls